Skip to content

AI / GenAI Engineer · Full-Stack Developer

Hamim Saad Al Raji

Building AI that ships, and research that holds up.

Final-year Software Engineering student at Islamic University of Technology, blending rigorous academic research with production-grade full-stack development.

Drawn by
Hamim Saad Al Raji
Location
Dhaka, Bangladesh
Scale
1:1
Status
Open to AI/ML roles & research
Academics
B.Sc. Software Engineering · CGPA 3.60 / 4.00
Sheet
00 · Cover
Rev
2026

Key data

10K+
Users served in production
IEEE 2026
Publication accepted
4
LLM-powered apps shipped
2nd Place
Techathon Nationals 2026

Selected Work

DWG. 01

AI Interviewer

An AI voice-interview platform: VAPI voice agents drive dynamic, LLM-generated questions with adaptive difficulty and speech-to-text transcription; a Firebase pipeline scores candidates via LLM-as-a-Judge on technical accuracy, communication, and problem-solving. Deployed on Vercel with serverless API routes.

Issued

2025

Bill of materials

  • NEXT.JS
  • VAPI
  • GEMINI
  • FIREBASE

References

DWG. 02

Robotic Arm — Simulation & Control System

A browser-based simulation and control system for a 6-DOF industrial robotic arm, built for the Techathon Nationals & Rover Summit 2026 finals — Runners-Up (2nd place). Real-time robot visualization with inverse kinematics, keyboard / joystick / voice control and an autonomous PIN-entry mode, extended with a Google Gemini natural-language interface.

Issued

2026

Bill of materials

  • SIMULATION
  • INVERSE KINEMATICS
  • VOICE CONTROL
  • GEMINI

References

DWG. 03

Campus Haat

A multimodal AI marketplace for universities: Gemini Vision auto-generates product titles and descriptions from seller photos, and Groq-hosted speech-to-text supports voice listings in English and Bangla. Leaflet-based meetup pinning uses campus geofencing as a default safety zone. Dockerized full stack, deployed on Vercel.

Issued

2025

Bill of materials

  • NEXT.JS
  • MONGODB
  • GEMINI VISION
  • GROQ
  • LEAFLET
  • DOCKER

References

DWG. 04

MarketPulse — Financial Sentiment Classifier

A fine-tuned DistilBERT model that labels financial sentences as positive, negative, or neutral. Benchmarked against a TF-IDF + Logistic Regression baseline to prove the transformer earned its cost — reaching 0.85 weighted F1 vs 0.76 — while staying fast at 14.86 ms per request. Published to the Hugging Face Hub and served via a FastAPI REST API, with a pytest suite and a hardened Docker image.

Issued

2025

Bill of materials

  • PYTORCH
  • TRANSFORMERS
  • FASTAPI
  • DOCKER
  • HUGGING FACE

References

DWG. 05

YouTube-Buddy — AI Chrome Extension

A Chrome extension with a Retrieval-Augmented Generation pipeline that lets users ask context-aware questions about any YouTube video, using its transcript as a knowledge base. Retrieval built with LangChain text splitters and vector embeddings; Google Gemini as the generation LLM; Python backend on Microsoft Azure with response caching.

Issued

2024

Bill of materials

  • LANGCHAIN
  • GEMINI
  • PYTHON
  • AZURE

References

DWG. 06

NijerJomi — Blockchain Land Registry

A blockchain-based land registry for Bangladesh ensuring tamper-proof, transparent property-ownership records. Finalist at Bangladesh Blockchain Olympiad 2025 among 200+ teams. Type-safe, responsive React + TypeScript frontend with shadcn-ui and Tailwind CSS for the land-transfer workflow.

Issued

2025

Bill of materials

  • REACT
  • TYPESCRIPT
  • VITE
  • SHADCN-UI
  • TAILWIND

References

Research

ACCEPTED
FIG. 1
100500
98.17
97.81
97.12
AccuracyPrecisionF1
Reported metrics on the IQ-OTH/NCCD CT test set — CoAtNet, held-out evaluation (%).

98.17%

Diagnostic Accuracy (Overall)

Evaluating CoAtNet for Multiclass Lung Cancer Classification on CT Images: A Benchmark Study on the IQ-OTH/NCCD Dataset

A benchmark of CoAtNet — a hybrid convolution–transformer architecture — for multiclass lung-cancer classification (benign, malignant, normal) on the IQ-OTH/NCCD CT-scan dataset, with selective augmentation, high-resolution resizing, and adaptive color-space preprocessing; evaluated against baseline CNN and ViT models.

IEEE QPAIN 2026 · CO-AUTHOR · PAPER ID 1101

S.E.S. Khan, K.H. Ador, H.M. Alvee, H.S. Al Raji, S.M.S. Alam, M.F. Mollah

Experience

Issued

OCT 2025 — FEB 2026 · Dhaka, Bangladesh

Full-Stack Developer Intern

Trust Innovation Limited — ICT Company of Bangladesh Army

10,000+
Records Handled

Revision history

RevDescription
  1. Led development of the TIL Education Management System (TIL-EMS) serving 10,000+ student records, delivering end-to-end full-stack features with React, TypeScript, Node.js, Express, and Prisma.

  2. Designed and deployed a financial automation ecosystem — a hierarchical Chart of Accounts, a double-entry voucher system, and a cron-based monthly fee engine that auto-generated and validated thousands of transactions.

  3. Refactored a 2,300-line monolithic module into a modular architecture; built engines for routine-conflict detection and automated result processing, plus secure payment-gateway integration. The platform is deployed across Army-run schools and colleges in three cantonments.

About

Final-year Software Engineering student at IUT focused on AI/GenAI and computer vision. I build LLM-powered, full-stack applications and work on deep-learning research for medical imaging. I'm looking for AI/ML engineering roles and funded research opportunities.

Download CV
4.1

Languages

  • Python
  • JavaScript
  • TypeScript
  • SQL
  • Java
  • C
  • C++
  • C#
4.2

AI / GenAI

  • LangChain
  • RAG
  • Embeddings
  • Vector Search
  • Prompt Engineering
  • Gemini/Groq APIs
  • VAPI
  • Model Context Protocol (MCP)
  • LLM-as-a-Judge
4.3

Deep Learning & Data

  • PyTorch
  • CoAtNet
  • Computer Vision
  • Medical Image Classification
  • NumPy
  • Pandas
  • scikit-learn
4.4

Frameworks / Infra / Cloud

  • React
  • Next.js
  • Node.js
  • Express
  • Prisma
  • FastAPI
  • Tailwind CSS
  • shadcn-ui
  • Docker
  • Git
  • GitHub
  • Azure
  • Vercel
  • Linux
  • MongoDB
  • PostgreSQL
  • Firebase
4.5

Education

  • Islamic University of Technology

    B.Sc. Software Engineering · CGPA 3.60/4.00

    Aug 2022 – Present

  • Dhaka College

    Higher Secondary Certificate · GPA 5.00/5.00

4.7

Spoken Languages

  • English — IELTS Academic 8.0 (C1)
  • Bengali — Native

Off the Clock

Life outside the terminal.

When I'm not shipping code or training models, I share everyday life and the student-builder grind on YouTube.

Watch on YouTube

Recognition

  • Runners-Up (2nd Place) — Techathon Nationals & Rover Summit 2026, IUT Robotics Society

  • Finalist — Bangladesh Blockchain Olympiad 2025 (team "Crypto Jamidar"; top finalist among 200+ teams)

  • Top 10 — CodeRush 2.0 Hackathon, IUT (2025)

  • IELTS Academic — Band 8.0 (C1 Advanced)